Method
What the bot actually does.
No part of this is a score projection. The whole system reads one thing, the gap between where the money is and where the tickets are, and bets that the number keeps moving.
The signal
Money against tickets
Sportsbooks report two different shares on a market: the percentage of bets on each side, and the percentage of money. When those two disagree sharply, a small number of large bets are on one side and a large number of small bets are on the other. The book screens for that gap, and for how far the price has already moved from its opening number. A pick needs both. A crowd it disagrees with but a line that has not moved is not a signal; a line that has moved for a reason everyone can already read, such as a starter ruled out or a goalie confirmed, is not one either.
Why closing-line value is the headline number. Fair value here comes from the same prices the book bets into, so the entry sits at roughly fair-minus-vig and there is no separate estimate of who should win. That makes one test the right one: did the market move the way the pick said it would. It is a harder number to flatter than a win rate, and it is on every page of this site.
Prices and grading
Where the numbers come from
Entry prices are FanDuel on almost every pick. Two exceptions worth naming: a little over twenty football and hockey picks are priced on Polymarket, and the NBA board records whichever book had the best number at the time, which is part of why the NBA page is measured in points rather than percent. Closing prices are captured before the game starts, never looked up afterwards. A pick with no captured close is shown with a dash rather than an estimate.
On the crowned pick. One pick a day is flagged as the best play, ranked on price action alone. Two things are being watched rather than acted on: most crowns so far rest on a divergence score that is largely structural rather than a read of the crowd, and the highest-scoring band has earned the weakest closing-line value in the book while winning the most games. The sample is small and the rule is frozen until it can be reviewed fairly.
CLV is reported in percentage points of implied probability: the implied probability of the closing price minus the implied probability of the entry price. A positive number means the market finished shorter than the price the pick was written at.